fix: allow ccd residues with missing coords

[davegrays]

There are a number of ligands/residues in the CCD with missing coordinates that cannot be retrieved now with structure.info.residue due to a ValueError that is raised twice and not excepted the second time. Relative to v0, this is new behavior.

This allows these residues to be returned from the second attempt with nans where missing.

[codspeed-hq]

CodSpeed Performance Report

Merging #730 will not alter performance

_{Comparing davegrays:fix/allow-all-ccd-residues (4e931fb) with main (226f2b4)}

Summary

✅ 59 untouched benchmarks

[padix-key]

Hi, from my perspective having a molecule without atom coordinates has barely a benefit. Consequently, getting an exception, that can be easily captured, is more clearer than checking for NaN coordinates in my opinion. However, I would be curious what your use case is.

[davegrays]

@padix-key Makes sense. My use case is that I'm identifying missing atoms in the experimental structure by constructing molecular templates from the CCD (essentially just the list of atom names per residue). So for that I use the CCD's AtomArray and I'm ignoring the coords anyway.

I get your concern tho... what would you think about an optional arg for this in get_components() that can propagate to structure.info.residue?

[Croydon-Brixton]

I agree with @davegrays that optionally exposing missing atoms in the builtin CCD is useful for building template residues of 'what should be there if all atoms were resolved' and thereby allows checking for missing atoms.

I like the idea of exposing a flag in get_components for this -- here 's a patch MR for this davegrays#2

[davegrays]

Thanks @Croydon-Brixton ! Merged your patch

@padix-key I implemented the remaining fix to propagate the allow_missing_coords flag to structure.info.residue. Curious what you think now.

[padix-key]

Hi, sorry for the delay. I see the point now 👍.

The PR looks mostly good to me, what misses is only a test in tests/structure/test_info() where residue() loads a molecule with missing coordinates.
Apart from that I added some small code style suggestions to the review.

[davegrays]

@padix-key thanks for your feedback! all the style changes are implemented and I've added test cases for a few different molecules

[davegrays]

also not sure exactly why but it seems like the ci/cd step that failed on this last commit was an intermittent network-related issue. i only changed the test itself and i can confirm the test still passes

[padix-key]

I just run it again, and check if it passes. But I agree, this should have nothing to do with your changes

[padix-key]

The CI now passes and everything looks good to me. Thank you!